12 rare event detection and propagation in wireless sensor ......rare event detection and...

12

Rare Event Detection and Propagation in Wireless Sensor Networks

DAVID C. HARRISON, Victoria University of WellingtonWINSTON K.G. SEAH, Victoria University of WellingtonRAMESH RAYUDU, Victoria University of Wellington

Rarely occurring events present unique challenges to energy constrained systems designed for long termsensing of their occurrence or effect. Unlike periodic sampling or query based sensing systems, longevitycan not be achieved simply by adjusting the sensing nodes’ duty cycle until an equitable balance betweendata density and network lifetime is established. The low probability of occurrence and random nature ofrare events makes it difficult to guarantee duty cycled battery powered sensing nodes will be energised whenevents occur. Equally, it is usually considered impractical to leave the sensing nodes energised at all timesif the network is to have an acceptably long operational life. In the past decade and a half, wireless sen-sor network research has addressed this aspect of rare event sensing by investigating techniques includingsynchronised duty-cycling of redundant nodes, passive sensing, duplicate message suppression and energyefficient network protocols. Researchers have also demonstrated the efficacy of harvesting energy from theenvironment to extend operational life. Here we survey existing rare event detection and propagation tech-niques, and suggest areas suitable for continued research.

Categories and Subject Descriptors: C.2.2 [Computer-Communication Networks]: Network Protocols

General Terms: Design, Algorithms, Performance, Reliability

Additional Key Words and Phrases: Wireless sensor networks, rare events, duty cycling, energy harvesting

ACM Reference Format:David C. Harrison, Winston Seah and Ramesh Rayudu, 2014, Rare Event Detection and Propagation inWireless Sensor Networks. ACM Comput. Surv. 0, 0, Article 12 ( 0000), 23 pages.DOI:http://dx.doi.org/10.1145/0000000.0000000

1. INTRODUCTIONTraditional wireless sensor network (WSN) nodes are small battery powered devicestypically consisting of a micro controller, a modest quantity of random access mem-ory, some non-volatile storage capacity, one or more sensors, and a low power radiotransceiver. The finite charge storage capacity of batteries shapes WSN research tothe extent that minimising energy consumption becomes a preoccupation; the less en-ergy consumed, the longer the network will continue to operate.

The simplest way for a sensing node to conserve energy, and in doing so maximiseits operational life, is to power down for extended periods. For star topology networkswhere sensing nodes are connected to a permanently powered base station, each nodecan adopt an independent duty cycle and a media access control (MAC) protocol basedon un-slotted carrier sense multiple access with collision avoidance (CSMA/CA). Nodesdeployed to periodically sample data for multi-hop transmission to a base station cansynchronise their activity [Ye et al. 2004] to ensure network connectivity and employalgorithms that minimise overuse of individual routing nodes [Schurgers and Srivas-tava 2001]. Similar techniques can be adopted for networks where data collection is ini-tiated by a request from the base station [Yao and Gehrke 2003]. Sensing rare events,however, introduces additional complexity.

Rare events are axiomatically situations that occur infrequently; they may also beshort-lived, present themselves unpredictably and leave no trace of their presencewhen complete. For the purposes of this work, rare events are considered random asthose following a predictable schedule can easily be sensed assuming accurate timesynchronisation [Elson 2003]. Successfully sensing a rare event requires considera-tion of the extent to which it is ephemeral and transitory. A scheme proved effectivefor sensing forest fires in progress may not be as successful detecting the instanta-

ACM Computing Surveys, Vol. 0, No. 0, Article 12, Publication date: 0000.

12:2 D. C. Harrison, W. K. G. Seah, R. Rayudu

neous start of the same fires. Similarly, a scheme for detecting perimeter intrusion ona battle field where events last fractions of a second and leave no discernible tracemay prove inappropriate for sensing landslides that last for comparatively extendedperiods and leave significant physical evidence in their wake, yet occur less frequently.

A study of rare event simulation using Monte Carlo Methods [Rubino and Tuffin2009] compliments an earlier work on estimating rare event probabilities [Glassermanet al. 1999] and defines a rare event as:

“an event occurring with a very small probability, the definition of ‘small’ dependingon the application domain”

Examples given include a civil aircraft failing during a typical eight hour flight anda high speed network node experiencing a buffer overflow. Another introductory texton rare event simulation [Bucklew 2013] does not propose a definition, concentratinginstead on the randomness and uncertainty of rare events.

It should be noted that the term ‘rare event’ is uncommon in WSN research, theliterature containing only a handful of papers with the words ‘rare’ and ‘event’ in theirtitle [Misra et al. 2015; Harrison et al. 2015; Xu et al. 2012; Cheon et al. 2009; Asur andParthasarathy 2007; Dutta et al. 2005; Cao et al. 2005a], of which only four have beencited by others1 [Cheon et al. 2009; Asur and Parthasarathy 2007; Dutta et al. 2005;Cao et al. 2005a]. However, we include in this survey research relating to event sensingin a more general sense that contains techniques, algorithms and analysis suitable forWSNs deployed to detect rarely occurring events.

Rare event WSNs are found in a variety of situations including the battlefield [Aroraet al. 2004] where low unit costs allow high density, short time-frame, disposable de-ployments. In industrial settings [Low et al. 2005] WSNs are cost efficient when com-pared to fixed wiring, where robust self-organizing characteristics make them suitablefor monitoring hazardous machinery and protecting high value assets. It is entirelypractical to perform periodic data collection and rare event detection in one WSN, butquality of service (QoS) aware routing protocols [Gelenbe and Ngai 2008] should beimplemented to prevent unacceptable delays when urgent message relating to rareevents have to queue up behind more mundane traffic. Whatever the deployment sce-nario, two metrics emerge as fundamental to rare event sensing [Liu et al. 2006; Caoet al. 2005a; Cao et al. 2005b]:

Detection Probability. Likelihood of an event being detected.Detection Delay. Time taken for notification to reach a network sink.

WSNs can be considered part of the Internet of Things (IoT) [Atzori et al. 2010] andcontinue to benefit from active research projects and regular publication of surveystargeting specific topics. Recently, statistical, probabilistic, artificial intelligence, andmachine learning methods for event detection have been briefly surveyed [Nasridinovet al. 2014] yet the majority of WSN surveys in the last five years have focused onareas unrelated to rare events. One exception [Pantazis et al. 2013] does consider eventsensing as it relates to energy efficient routing with an emphasis on how the criticalityof an event informs protocol design. Table I lists ten of the most frequently cited WSNsurvey papers published since January 1, 2010 with their area of interest.

This paper is, to the best of our knowledge, the first to survey WSN research thateither specifically targets rare event sensing, considers work on rare events to be re-lated, or documents ways the presented work could be applied to rare event sensing.Issues generic to all WSNs, including but not limited to security [Walters et al. 2007],

1As reported by Google Scholar, August 19, 2015


Rare Event Detection and Propagation in Wireless Sensor Networks 12:3

Table I: Most cited WSN Surveys since January 2010

Survey Area of Interest

[Cheng et al. 2012] Localisation

[Sundani et al. 2011] Simulators

[Zhang and Varadharajan 2010] Key management

[Katiyar et al. 2011] Heterogeneous clustering algorithms

[Sen 2010] Security

[Rodrigues and Neves 2010] IP based WSNs

[Fan and Jin 2010] Sensing coverage

[Christin et al. 2010] Industrial automation

[Martins and Guyennet 2010] Security

[Mohanty et al. 2010] Security

software distribution [Han et al. 2005], and radio interference [Zhou et al. 2005] aredeemed out of scope. Similarly, this survey does not consider event sensing techniquesunrelated to the rarity of the event in question. Table II illustrates the relative volumeof surveyed work for the following event sensing strategies:

Collaboration. In large multi-hop sensing networks, the energy cost of long-distance communication is significant. If sensing nodes work locally to collabora-tively remove false positive event detections, network lifetime can be extended with-out adversely impacting detection probability. Collaboration can also take the formof nodes with discrete capabilities working together to more reliably detect rareevents.Component Deactivation. It is not always necessary for all WSN node compo-nents need to be constantly energised. Deactivating power hungry devices, partic-ularly radio transceivers, saves energy and leads to longer network life-spans, ex-tending the period in which a high detection probability can be maintained.Duty Cycling. Shutting the entire sensing node down periodically is the ultimateenergy saving technique. When nodes synchronise their sleep schedules, predictablelevels of detection probability and delay can be maintained whilst extending the lifeof the network.Over Population. Densely deploying sensing nodes such that they introduce re-dundancy in the network enables local collaboration and synchronised duty cyclingin addition to an increased tolerance to individual node failure.Message Suppression. When a number of sensing nodes detect the same rareevent, their detection messages may be regarded as duplicates of each other. Sup-pressing these duplicates reduces network traffic and saves energy in doing so.Burst Aware Protocols. When a rare event is detected by multiple sensors, aburst of network traffic is generated calling for protocols designed to handle theresulting media contention without introducing unnecessary detection delay.Always On. Whilst not suitable for all deployment scenarios, the simplest strategyfor maximising detection probability and minimising detection delay is to perma-nently energise all nodes in the network.Energy Harvesting. The challenge of maximising network longevity and detec-tion probability, and minimising detection delay can be moderated by harvestingenvironmental energy to augment or replace batteries in sensing nodes.



Table II: Volume of Surveyed WSN Research on Rare Events

Sensing Strategy Papers Surveyed

Collaboration 17

Duty Cycling 15

Component Deactivation 14

Over Population 8

Message Suppression 7

Burst Aware Protocols 7

Always On 7

Energy Harvesting 6

The body of this paper is organised as follows: Section 2 details commonly usedstrategies for rare event sensing and surveys existing research, Section 3 discussesissues arising from this survey and Section 4 briefly considers work in associated fields.Conclusions are drawn and open issues are highlighted in Section 5.

2. EVENT SENSING STRATEGIESDiversity in the characteristics of rare events prevents any one sensing strategy frombecoming a panacea. Further, even given a well defined event classification, utilizinga single strategy may be insufficient to minimise occurrences of undetected eventswhilst maximizing the operational life of the sensing network. This section introducesevent sensing strategies with examples of existing work examining them and surveysto what extent each strategy is represented in the literature. Table III summarisessurveyed event sensing strategies, indicating their typical impact on detection prob-ability and detection delay. The following sub-sections describe each strategy in moredetail.

2.1. CollaborationIn large multi-hop networks, the cost of propagating event detection messages fromoriginating nodes to the base station can be significant. If adjacent nodes collaborateprior to initiating the expensive long distance communication, a group decision is madeon whether or not sufficient collective sensing evidence exists to support publication ofan event detection message. A survey of cooperative event detection algorithms [Wit-tenburg et al. 2012] claims distributed evaluation exhibits the best energy efficiencyand highest detection accuracy when compared to local and centralised approaches.

In a recently proposed probabilistic event monitoring scheme (PEMS) for sparsenetworks [Das and Misra 2015], sensing nodes detecting an event collaborate to se-lect those whose observations are probabilistically significant. Nodes collaborativelyselected for event monitoring then adjust the network topology for improved energyefficiency by assigning responsibility for data forwarding to a single clusterhead.

A system capable of being trained to recognise application specific event types [Wit-tenburg et al. 2010] has been field-tested for construction site intrusion detection with100 nodes. Transmitting sensed data to a base station for analysis is eschewed infavour of collaboration between sensing nodes to determine if an event has occurred.Each node has multiple sensors and extracts feature vectors from raw data. Adjacentnodes share their feature vectors; if and only if they aggregate to match trained orstatistically defined event vectors will a detection message be published.



Table III: Typical Effects of Sensing Strategies

Detection

Strategy Probability Delay Observations

Collaboration Increases Increases

Collaborative sensing can lead to a more accurate un-derstanding of when rare events have occurred andwhat does and does not constitute a false positive de-tection. However, the time taken for nodes to exchangecollaboration information inevitably delays transmis-sion of event notification messages.

ComponentDeactivation - Increases

The lack of effect on detection probability assumes pri-mary sensors are not the components being deacti-vated. Powering off the transceiver inevitably leads totransmission delays and in some circumstances mayalso increase overall energy consumption as poweringup the transceiver can consume as much energy astransmitting one packet [Olds and Seah 2012].

Duty Cycling Reduces Increases

When combined with an over population of nodes, neg-ative impact on detection probability can be reduced,yet control messages for synchronised schemes intro-duce network overhead.

OverPopulation Increases May

Increase

Whilst a higher density of sensing nodes is beneficialto event detection, the additional network complexitycan lead to delays in getting the message out.

MessageSuppression - Decreases

Removal of duplicate messages removes congestionfrom the network allowing urgent event notificationsto reach their destination more rapidly.

Burst AwareProtocols - Decreases

Protocols specifically designed to handle a flurry ofnetwork activity are primarily aimed at reducing de-tection delay. Such protocols typically have a compo-nent deactivation element which needs to be carefullyconfigured to minimise the increase in detection delaywhilst waiting for deactivated components to warm upagain.

Always On Maximises MinimisesStored energy is rapidly exhausted making long termdeployments problematic. Combination with energyharvesting promises extension of network life.

EnergyHarvesting Increases -

Networks powered by energy harvesting have the po-tential to stay active for longer than those relying onbatteries. However, energy availability may be unpre-dictable leading to temporary reductions in detectionprobability caused by unexpected node outages.

Rather than a simple threshold breach, more sophisticated event signatures canbe used for distributed detection of transient complex events. In [Martincic andSchwiebert 2006] nodes are assumed to be stationary, location aware and uniformlydistributed, with a path existing between each pair of nodes. The sensor network ispartitioned into equal size cells. Nodes within a cell rotate leadership responsibilitiesincluding intra-cell communication where a time stamped weighted average value forthe cell is shared with adjacent cells. For a network represented by p×q cells, an eventsignature is an r × s matrix where r < p and s < q. To detect an event, leadershipnodes “overlay” the event signature on the nodes in their cell; if a match is found, theevent is considered to have been detected and an appropriate message is forwarded tothe base station.



Distributed Bayesian algorithms for detection of sensing faults in WSNs tasked withthe binary detection of environmental events [Krishnamachari and Iyengar 2004] havebeen shown through theoretical analysis and simulation to correct 85-96 percent offaults even when as many as 10 percent of the nodes are faulty. [Luo et al. 2006]present related simulation work whilst also evaluating Neyman-Pearson approachesand more closely considering the energy efficiency of their algorithms.

[Bandyopadhyay and Coyle 2003] identify that clustering nodes and assigning aclusterhead that takes responsibility for forwarding detection messages from the clus-ter to a base station may save energy. The authors propose a distributed, randomisedclustering algorithm and extend it to generate a hierarchy of clusterheads. Energyefficiency is observed to improve as the number of levels in the hierarchy increase.

Real-time event detection services are provided in Data Service Middleware(DSWare) [Li et al. 2003] by collaborative correlation of sensing observations based onevent characteristics enabling in-network differentiation between event occurrencesand false alarms. Data semantic based confidence functions are supported to deter-mine the relative importance of sub-events and capture historical patterns. If detectionrates are low, partial detections of critical events are reported.

Military field tests of a distributed target classification system based on collabora-tive signal processing [Brooks et al. 2003] proved successful. Objects in the sensor fieldgenerate time varying spatial signatures from multiple sensors, and a moving objectis a peak in the signature field that moves over time. Tracking the object is equivalentto tracking the peak. Distributed tracking is achieved by dividing the sensor field intocells within which a node is designated the manager, responsible for inter-cell coor-dination and intra-cell communication. In collaborative detection, classification, andtracking of moving targets: (1) cells near potential target trajectories are alerted andnodes in these cells collaborate to sense a target; (2) when a target is detected, thecell becomes active and tracking is initiated if the target is of the required type; (3)estimates of the target’s current location and velocity are used to estimate potentialfuture locations of the target; (4) as the target approaches the cell boundary, cells onthe estimated trajectory are alerted and the process repeats.

The Role Alternating Coverage Preserving Coordinated Sleep Algorithm (RACP)[Hsin and Liu 2004] and the Equitable Sleep Coverage Algorithm for Rare GeospatialOccurrences (ESCARGO) [Harrison et al. 2015] collaboratively modify the duty-cycle(Section 2.2) of an over population (Section 2.4) of sensing nodes to reduce network-wide energy consumption whilst maintaining sensing coverage.

Energy efficiency through energy awareness of cooperating sensor nodes is the focusof another surveillance system [He et al. 2004] aimed at military deployments. Tradeoffs between energy and surveillance efficiency are made by adjusting the sensitivity ofthe system. On initialization, nodes discover their neighbours and elect sentry nodeswho monitor the environment for intruders, while the remaining nodes enter a lowpower mode. If an intrusion is detected by a sentry, it wakes up the other nodes andthey collaborate to track the intruder. When the intruder leaves the sensing area, thesystem resets, alternate sentries are selected, and the process continues.

Collaborative fuzzy logic schemes for event detection have been proposed [Kapi-tanova et al. 2012; Thuc and Insoo 2011; Liang and Wang 2005] and surveyed [Sharmaand Singh 2014] though specific applications to rare events have yet to be documented.One challenge fuzzy logic algorithms face is that they typically require large rule-bases, the storage and distribution of which can prove problematic for the limited re-source devices typically used in WSNs. Less specific machine learning techniques haverecently been proposed for collaborative detection of definitively rare events, namelyleaks in water distribution pipelines [Rashid et al. 2014].



Works detailed in other sections of this survey that also feature collaborative eventsensing are: [Mekikis et al. 2013; Alam et al. 2012; Milic 2012; Arora et al. 2004; Yeet al. 2004].

2.2. Duty CyclingIn many event sensing scenarios, an always on strategy (Section 2.7) fails to meetlongevity requirements beyond a few days or weeks. For extended use, regularly pow-ering nodes down (putting them into sleep mode) improves operational life, potentiallyat the expense of guaranteed event detection.

Maximizing the detection probability for transient events and minimizing the detec-tion delay for persistent events are fundamental to acceptance of duty cycling as anappropriate strategy for event-driven WSNs. [Zhu et al. 2012] characterise the trade-off between system lifetime and detection performance; an algorithm to collaborativelydetermine when nodes should wake up is proposed and a favourable comparison ismade to random sleep scheduling.

Recent research [Misra et al. 2015] specifically targeting rare events proposes aprobabilistic duty cycle in sensor medium access control (PDC-SMAC) algorithm forinfrequent events in a military scenario intended to reduce the energy cost of ‘ineffec-tive sensing’ - periods where a node is energised but there is nothing to sense. Sleepscheduling [Kavitha and Lalitha 2014] promises bounded detection delay while maxi-mizing network lifetime. [Cao et al. 2005a] introduce a scheme where nodes enter andexit sleep mode in a coordinated fashion ensuring sensing coverage rotates over thenetwork, each node being energised at regular intervals. For persistent events, thisscheme introduces an increase in detection delay, with a reduction in detection proba-bility being inevitable for transient events. In scenarios where network longevity takespriority, schemes of this nature may be acceptable.

For situations where the WSN is responsible for alerting humans in the sensingarea, such as forestry workers when a fire breaks out or miners when a gas leak is de-tected, detection delay is considered the critical factor as the complex multi-hop natureof disseminating information through a large network can be adversely impacted byintermediary nodes adopting sub-optimal duty cycles. A level-by-level offset schedulebased duty cycle pattern [Guo et al. 2012] has been proposed that restricts the upperbound of detection delay to be 3D + 2L where D is the hop count to a central node andL is the duty cycle sleep time. The scheme proposed in [Guo et al. 2012] adopts a twophase alarm broadcast following detection of the rare (the authors use the term ‘crit-ical’) event. In the initial phase an event event notification is sent to a central nodewhich initiates the second phase where the alarm is broadcast to all other nodes.

Adaptive control of the duty cycle [Vigorito et al. 2007] for energy harvesting WSNs(Section 2.8) allows sensing nodes to respond to changes in the environment. Suchschemes have been shown to perform better than techniques that require a prioriknowledge of available energy. This combination of techniques typifies successful rareevent sensing research where two or more approaches compliment each other. Dutycycling and energy harvesting are individually advantageous techniques that whencombined provide even greater benefit.

Full node duty cycling is also considered in the following papers detailed in othersections of this survey: [Kang et al. 2012; He et al. 2006; Keshavarzian et al. 2006; Heet al. 2004; Hsin and Liu 2004; Kumar et al. 2004; Tian and Georganas 2002]

2.3. Component DeactivationAn alternative to full node duty cycling (Section 2.2) is powering down componentswhen they are not required. The component of WSN nodes that uses the most energyis the wireless transceiver (Table IV). Transmit and receive costs are inevitable, but



energy is wasted by media collisions and associated retries. Idle listening, where thetransceiver is powered up but receives no packets, is another significant waste of en-ergy. Deactivation of components can be application controlled or initiated by otherlayers, most notably MAC where the transceiver is often allowed to sleep when thereis nothing for it to listen for or receive.

AIMRP, an Address-light Integrated MAC and Routing Protocol [Kulkarni et al.2006] is specifically aimed at rare event detection scenarios. A randomised power sav-ing mode allows nodes to de-activate their transceivers independently of one another.Using simulations, AIMRP has been shown to compare favourably to the more genericsensor MAC (S-MAC) [Ye et al. 2002] for event detection applications. S-MAC also sup-ports transceiver deactivation, virtual clusters of nodes perform localised sleep syn-chronization. Energy consumed by AIMRP and S-MAC are compared for selected val-ues of τ , the maximum permissible end-to-end latency for event reports, τ being equiv-alent to the maximum allowable detection delay; Simulation and analysis of AIMRPshows it consumes less energy than an analysis of S-MAC suggests it would for valuesof τ between 0.015s and 0.05s.

At the application level, a military intrusion system “Line in the sand” (LITS) [Aroraet al. 2004] makes the distinction between passive and active sensing. Passive sensorsare ones that measure analogue properties of the intruder such as magnetic, ther-mal and acoustic characteristics, while active sensors are those that determine theintruders’ range, velocity and direction of travel by how the target modifies, reflects orscatters a signal transmitted by the sensor. The active sensor in LITS is pulse Doppler,while a magnetometer is used as a passive sensor. Detection of an intruder by the co-operatively low powered passive sensor triggers initialization of the active sensors. InLITS the active and passive sensors are on different nodes which collaborate, with thepassive nodes detecting the intrusion and the active nodes tracking it.

A source-initiated or sink-initiate wakeup radio (WUR) based medium access con-trol (GWR-MAC) [Karvonen et al. 2014] aims to reduce idle listening and by doingso improve energy efficiency in short range communication networks such as wirelesssensor and body area networks. Favourable analytical comparisons are made againstconventional duty-cycled (Section 2.2) MAC Protocols suggesting GWR-MAC would beuseful and energy efficient for sensing low frequency events requiring a reasonably lowdetection delay.

The TRafficAadaptive Medium Access protocol (TRAMA) [Rajendran et al. 2006] is atime slotted MAC protocol with a distributed election scheme. TRAMA guarantees col-lision free transmission and allows nodes to power down their transceivers when theyare not transmitting or required to receive. Additional MAC protocols that deactivatethe transceiver for short periods are described in Section 2.6.

Messages used to control a sensing network are typically modestly sized and sentin small batches at high frequency; those sent in response to event detection tend tobe much larger, sent in higher volumes but far less frequently. Deploying two radioson each sensing node [Feng and Potkonjak 2002], a low power device for control trafficand a higher power component for event data, allows the more energy hungry deviceto be powered down almost all the time.

Component deactivation is also considered in the following papers detailed else-where in this survey: [Yoo et al. 2012; Sun et al. 2008a; Sun et al. 2008b; Dutta et al.2005; He et al. 2004; Polastre et al. 2004; Ye et al. 2004; Van Dam and Langendoen2003]

2.4. Over Population / Node RedundancyIn a network with sufficient sensing coverage when all nodes are energised, duty cy-cling (Section 2.2) increases network longevity, but reduces coverage. Over populating



Table IV: Current Draw for WSN Components

Active Sleep

Texas Instruments CC2420 Transceiver1 18.8 mA 1 µA

Texas Instruments MSP430 Microcontroller1 1.8 mA 5.1 µA

Analog Devices ADXL345 3-Axis Accelerometer2 40 µA 0.1 µA

Panasonic EKMB123112 Passive Infrared Sensor3 2 µA -1 From www.memsic.com2 From www.analog.com3 From pewa.panasonic.com

the sensing area allows nodes to synchronise their duty cycle so sensing coverage ismaintained when some subset of nodes are powered down. Installing redundant nodeswhere each primary sensing node has a partner in close proximity allows optimal cov-erage to be maintained and can provide some tolerance to faults.

SenSlide [Sheth et al. 2005], a landslide prediction system, is a WSN hybrid of eventdetection and data capture. Energy-aware routing protocols are utilised to avoid indi-vidual nodes becoming drained of power prematurely but a level of fault tolerance isprincipally achieved by deploying redundant nodes.

For ephemeral events where a less than perfect detection probability is unaccept-able, duty cycling algorithms can preserve coverage when an over population of nodesis deployed. A coverage preserving role alternating algorithm [Hsin and Liu 2004] en-ables such a deployment to extend the period during which initial sensing coverage ispreserved by allowing nodes to enter sleep state if a minimal subset of neighbouringnodes have agreed to collaboratively take responsibility for the sleeping node’s sens-ing area. When the sleeping node wakes up, sponsoring neighbours are now free toenter sleep mode if they can agree on a sponsoring contract with a group of their ownneighbours.

Other works surveyed that rely on, or consider an over population of sensing nodesare: [Harrison et al. 2015; Zhu et al. 2012; He et al. 2006; Cao et al. 2005a; Kumar et al.2004; Arora et al. 2004; Tian and Georganas 2002]

2.5. Message Suppression & Data AggregationWhen a rare event is sensed by multiple nodes in close physical proximity, they gener-ate messages that can be considered duplicates of one another; delivery of a subset ofthese messages may be sufficient to confirm event occurrence [Heinzelman et al. 2002;Yang et al. 2013].

CC-MAC [Vuran and Akyildiz 2006], composed of an event MAC (E-MAC) and a net-work MAC (N-MAC), exploits the spatial correlation of messages generated by sens-ing nodes to suppress duplicates from nodes in close physical proximity. In a givencorrelation region, a single node is chosen as being representative of its correlationneighbours; only messages originating at the representative node are transmitted tothe network sink, all other messages are suppressed.

Sift [Jamieson et al. 2006] is a slotted MAC protocol designed to deal with boththe bursty nature of a large number of nodes wishing to transmit at the same time(Section 2.6) and de-duplication of the message queue. In removing some percentage ofduplicate messages, Sift aims for the collision free delivery ofRmessages fromN nodeswhere R < N . On successful delivery of R messages, the remaining N − R messagesare suppressed.



Message aggregation techniques for data-centric routing [Krishnamachari et al.2002] include suppression of duplicates to reduce energy usage by minimisingtransmission volumes. In dense networks, greedy aggregation has been shown [In-tanagonwiwat et al. 2002] to be more efficient than opportunistic aggregation schemes.Sliding window skylines [Borzsony et al. 2001] have been also proposed as a suitabletechnique for message suppression in a system for forest fire detection [Pripuzic et al.2008] .

2.6. Burst Aware ProtocolsRare events can trigger a tsunami of messages in the sensor network, all of whichwould like to be transmitted at the same time leading to channel contention, an in-creased probability of packet collision and the potential for data loss and delay.

The MAC layer defined by the IEEE 802.15.4 standard [Gutierrez et al. 2001] andused extensively in WSN devices has two modes of operation. In beaconed mode,at least one device acts as a personal area network (PAN) coordinator and non-coordinator devices must wait to transmit in contention free time slots. If beaconingis not employed, an un-slotted carrier sense multiple access with collision avoidance(CSMA/CA) algorithm based on listening to the physical medium is used, collisionsbeing avoided by invoking a random exponential back-off algorithm. Beaconed or not,IEEE 802.15.4 MAC is designed to minimise energy usage in the nodes, not maximisedelivery rates. In extremely bursty conditions, an alternate MAC may be required.

A geographical cross-layer asynchronous sender-oriented MAC protocol [Zayaniet al. 2014] has recently been proposed to facilitate opportunistic routing in low duty-cycle wireless sensor networks (Section 2.2) that may experience bursty traffic patternsand unpredictable changes in network topology. This protocols objective is to maxi-mize network lifespan while guaranteeing packet delivery with acceptable end-to-enddelays.

Reliable Bursty Convergecast (RBC) [Zhang et al. 2007] is a MAC protocol designedspecifically to handle bursts of data in multi-hop networks that converge on a limitednumber of sinks. Using data traces taken from [Arora et al. 2004] an implementation ofRBC based on B-MAC [Polastre et al. 2004] is shown experimentally to perform betterthan the default TinyOS [Levis et al. 2005] radio stack where 100% packet delivery isnot required.

S-MAC [Ye et al. 2004] is an earlier MAC protocol for ad-hoc deployments of batteryoperated sensor nodes that remain inactive for extended periods but become suddenlyactive when an event is detected. S-MAC values energy conservation and self configu-ration over per-node fairness and latency. Virtual clusters of nodes adopting commonsleep schedules reduce control overhead and message passing techniques can reducecontention latency for applications undertaking in-network data processing.

A Receiver-Initiated asynchronous duty cycle MAC protocol for dynamic trafficloads (RI-MAC) [Sun et al. 2008b] is a transceiver deactivation (Section 2.3) proto-col shown to outperform contemporaries when faced with bursty message flows. Duty-cycle Scheduling based on Residual energy (DSP) and Duty-cycle Scheduling basedon Prospective increase in residual energy (DSR) [Yoo et al. 2012] are protocols de-rived from RI-MAC that determine their deactivation schedules based on residual orprospective increases in harvested energy (Section 2.8).

A MAC protocol designed to cope with the burst of traffic that follows detectionof a rare event is Demand Wakeup MAC (DW-MAC) [Sun et al. 2008a]; sleepingtransceivers are woken on demand yet data transmissions are guaranteed not to col-lide at receiving nodes. DW-MAC is a synchronised duty cycle MAC protocol that as-sumes accurate clock synchronisation by. Each cycle is divided into three periods: Sync,



Data and Sleep. DW-MAC wakes up nodes on demand during the Sleep period of a cyclein order to transmit or receive a packet.

Sift [Jamieson et al. 2006] is a MAC specifically for event sensing networks, its prin-ciples being similar to those found in rare event WSNs powered by energy harvestedfrom the event itself [Cheng et al. 2013] where bursts of data must be dealt with beforethe limited available energy is consumed.

2.7. Always OnLeaving event sensing nodes powered on at all times may be necessary if missing anevent occurrence is unacceptable, notification of the event needs to be as fast as possi-ble or the network deployment timescale is sufficiently short that battery depletion isunlikely. Estimating the longevity of always on sensing networks is problematic. In astudy of a large intrusion detection system [Kumar et al. 2005] it is suggested that mis-takes as large as an order of magnitude are routinely made. Nevertheless, adopting analways on strategy remains an appropriate course of action in certain circumstances.

Structural engineers monitoring vibrations require rich data at high sample ratesyet bandwidth limitations in WSNs made up of more than a handful of nodes makereliable delivery of such a data deluge impractical. Wisden [Xu et al. 2004] observesthat a single triple axis accelerometer generating 16-bit samples at 100Hz requires4.8Kbps and suggests delivery of data relating to interesting events (rather than acontinuous time series) is sufficient for analysis of structural vibration. Even takingthis into consideration and allowing for data compression before transmission, in suchan environment sensor nodes are busy almost all the time.

Monitoring events on an active volcano with small low-power WSN devices [Werner-Allen et al. 2006] was achieved by leaving the sensing nodes energised at all times.The short deployment (three weeks), volume of events detected (230), their ephemer-ality (less than 60 seconds each) and transient nature made an always on strategyacceptable. Similarly, a system for surveillance of temporary museum exhibits [Vianiet al. 2012] stays energised at all times to maximise detection probability; extendingnetwork life is of little concern as batteries can provide sufficient power for an entireexhibition.

Permanently powering all WSN nodes is a technique adopted in a military intru-sion detection system [Arora et al. 2004] and is referenced when considering coveragemaintenance in low duty cycled WSNs [Hsin and Liu 2004].

2.8. Energy HarvestingWSN research has traditionally focused on minimizing energy consumption in bat-tery powered sensor nodes, balancing longevity against functionality. Increasingly, re-searchers are evaluating the efficacy of harvesting energy from a variety of environ-mental sources [Gilbert and Balouchi 2008; Sudevalayam and Kulkarni 2011].

Energy sources such as solar are reliable (the sun always comes out) but unpre-dictable (it may be obscured by clouds). For rare event sensing where the event mayoccur at a point where limited environmental energy is available, a Harvest-Store-Usearchitecture [Sudevalayam and Kulkarni 2011] coupled with multiple storage devicesas used in Prometheus [Jiang et al. 2005] and AmbiMax [Park and Chou 2006] wouldappear appropriate, yet neither Prometheus nor AmbiMax address rare event sensingdirectly, Prometheus being specifically aimed at periodic data capture.

A proposed adaptive scheduling scheme for cooperative energy harvesting networks[Ammar and Reynolds 2015] combines cooperative communications (Section 2.1) andenergy harvesting in a scheduling scheme intended to maximizes packet delivery ratiowith the aim of gaining acceptable detection probability without necessarily minimis-ing detection delay. Cooperating nodes advertise their status as either ‘active/on’ or



’inactive/off ’ (a form of duty cycling (Section 2.2) but without component deactivation(Section 2.3)) based on their harvested energy levels. When a node detects an event andwishes to transmit an notification message, if a cooperating node is currently markedas ‘active’, the source node will forward the message via it’s cooperating neighbour;otherwise the source nodes attempts to transmit on its own. Simulation results indi-cate the scheme would provide similar performance to a state-of-the-art alternative [Liet al. 2012], but does not require threshold parameter optimization.

An adaptive scheme for duty cycling (Section 2.2) [Vigorito et al. 2007] in energyharvesting WSNs does not attempt to model the energy source allowing usage in sit-uations where a prior knowledge of the source is unavailable. The scheme’s compu-tational efficiency and adaptability to near depletion scenarios prevent node outageswhilst tunable stability restricts variance in individual node duty cycles. Similarly,ODMAC [Fafoutis and Dragoni 2011], an on-demand MAC specifically aimed at en-ergy harvesting WSNs supports individual duty-cycles for nodes with different energyprofiles but does not consider its application to rare event sensing.

DSR and DSP [Yoo et al. 2012] (cf: Section 2.6) propose dynamic duty cycle schedul-ing for energy harvesting WSNs based on RI-MAC [Sun et al. 2008b]. DSR allowssensor nodes to adjust their duty cycle based on their remaining energy store. DSPestimates future energy harvesting opportunities and aggressively adjusts duty cyclesproportionately. Reductions in end-to-end delay are demonstrated for DSR and DSPover unmodified RI-MAC for N ×N grids, when N is in the range 4m to 8m, with 20mbetween each node and a transmission range of 30m.

A survey of energy harvesting for structural health monitoring (SHM) sensor net-works [Park et al. 2008] did not restrict itself to rare events but recent research [Chenget al. 2013; Tomicek et al. 2013] addresses SHM via WSNs powered by energy extractedfrom the rare event itself, namely an earthquake. The focus is on a MAC protocol thatis capable of handling both the bursty nature of the messages generated by the event(Section 2.6) and the limited energy extracted from the earthquake via frequency tunedpiezoelectric vibration energy harvesters.

A detailed survey of power management in energy harvesting WSNs [Kansal et al.2007] considers event monitoring as a deployment example. Particular attention ispaid to detection delay making a distinction between the delay introduced by the MAClayer and energy aware routing protocols.

2.9. Classification of Existing WorkTable V summarises work discussed in this section highlighting which sensing strate-gies are featured in deployed applications, MAC protocols and supporting technologies.



Table V: WSN Research on Rare Event Sensing

Duty Cyc

ling

Compo

nent

Deacti

vatio

n

Over Po

pula

tion

Energ

y Harve

stin

g

Collab

orat

ionBur

stAwar

e Proto

cols

Mes

sage

Supp

ress

ion

Alway

s On

Applications[Mekikis et al. 2013] x

[Cheng et al. 2013] x x[Tomicek et al. 2013] x

[Viani et al. 2012] x[Wittenburg et al. 2010] x

[He et al. 2006] x x[Werner-Allen et al. 2006] x

[Dutta et al. 2005] x[Sheth et al. 2005] x[Arora et al. 2004] x x x x

[He et al. 2004] x x x[Xu et al. 2004] x

MAC Protocols[Zayani et al. 2014] x x

[Karvonen et al. 2014] x[Yoo et al. 2012] x x x

[Sun et al. 2008b] x x[Sun et al. 2008a] x x

[Rajendran et al. 2006] x[Vuran and Akyildiz 2006] x x

[Jamieson et al. 2006] x x[Polastre et al. 2004] x

[Ye et al. 2004] x x x[Van Dam and Langendoen 2003] x

[Ye et al. 2002] xSupporting Technologies

[Ammar and Reynolds 2015] x x[Tang et al. 2015] x

[Misra et al. 2015] x x[Das and Misra 2015] x[Harrison et al. 2015] x x x x

[Rashid et al. 2014] x[Kavitha and Lalitha 2014] x

[Yang et al. 2013] x[Zhu et al. 2012] x x

[Milic 2012] xContinued on next page. . .



Table V – Continued from previous page

Duty Cyc

ling

Compo

nent

Deacti

vatio

n

Over Po

pula

tion

Energ

y Harve

stin

g

Collab

orat

ionBur

stAwar

e Proto

cols

Mes

sage

Supp

ress

ion

Alway

s On

[Alam et al. 2012] x[Guo et al. 2012] x

[Kang et al. 2012] x[Thuc and Insoo 2011] x

[Wittenburg et al. 2010] x[Pripuzic et al. 2008] x[Vigorito et al. 2007] x x

[Martincic and Schwiebert 2006] x[Keshavarzian et al. 2006] x

[Luo et al. 2006] x[Cao et al. 2005a] x

[Kumar et al. 2005] x[Hsin and Liu 2004] x x x

[Krishnamachari and Iyengar 2004] x[Kumar et al. 2004] x x

[Bandyopadhyay and Coyle 2003] x[Li et al. 2003] x

[Brooks et al. 2003] x[Feng and Potkonjak 2002] x

[Heinzelman et al. 2002] x[Intanagonwiwat et al. 2002] x[Krishnamachari et al. 2002] x

[Tian and Georganas 2002] x x

3. DISCUSSIONWSNs can be classified by their data delivery profile as being continuous, event-driven,observer-initiated (query-based) or a hybrid [Tilak et al. 2002]. Based on the literature,it appears continuous, periodic sensing is the model that has been of most interest toWSN researchers, though event-sensing is increasing in popularity. The most widelycited1 survey on WSN research [Akyildiz et al. 2002] briefly mentions event sensingand query based approaches, but the majority of the technologies and systems it sur-veys target periodic sensing deployments. A more recent survey [Yick et al. 2008] givesmore space to event sensing and a survey of the state of the art of WSN programmingtechniques published within the last three years [Mottola and Picco 2011] features theapplication of event-triggered distributed processing.

1Citations as of October 7, 2015 - Google Scholar: 14,097, Microsoft Academic Search: 4,249.



Rare events are characterised by domain specific low occurrence probabilities, butselecting appropriate detection and propagation techniques relies on an understatingof a number of other attributes:

Frequency. While the probability of a rare event occurring in a given time periodmay be low, over an extended time period occurrences may become regarded asfrequent. For example, the probability of an earthquake hitting New Zealand inthen next given sixty seconds is low, yet the country experiences several hundredearthquakes a year.Randomness. To what extent do the events occur in a partially predictable fash-ion? In industrial settings there may be a trigger event such as certain piece ofmachinery being switched on that makes the rare event possible, but not necessaryprobable.Ephemerality. How long a rare event lasts can inform strategies for dealing withit. If it an event typically lasts for n time units, is it acceptable to detect in the nthtime unit? To what extent is it better to detect it in the 1st time unit?Transitory Nature. After an event occurs, is there any evidence it happened? Ifthere is, is it acceptable to detect the effect rather than the event? A fracture in awater main can be sensed either as a physical change at the location of the break(potentially problematic for a long pipe) or as a drop in water pressure beyond thebreak.Connectedness. Once an event has occurred, does that imply something about thenext occurrence? Can the magnitude or timing of subsequent events be predicted?Measuring the ephemerality of events as they occur could inform a sleep-schedulingalgorithm that starts by assuming events are instantaneous but learns that theevent always last (i.e. are detectable) for some period of time, hence the sleep cyclecan be modified to be slightly less than that period.Criticality. How important is it that an event be detected? Once detected, howquickly must a notification message be propagated to interested parties? This maybe related to event frequency, it may be acceptable to miss frequently occurring lesscritical events simply because it is a certainty another one will occur in the nearfuture.Network Lifetime. How long must this network be active? Is the event so rarethat once detected the network is permitted to effectively destroy itself by using allits energy to propagate the notification message.

To detect a rare event at a given location, at least one node within sensing range ofthat location must be operational when the event occurs. Similarly, once the event hasbeen detected, sufficient nodes must be active such that a low latency route exists forforwarding event notification messages to the network sink. Whether powered by bat-teries or energy harvested from the environment, WSNs deployed to detect and reportrare events are constrained by the twin requirements of maximising detection prob-ability and minimising detection delay whilst working with a limited energy source.The event sensing strategies surveyed in this paper address these constraints, eitherindividually or in complimentary combinations.

Two themes are highlighted in Table V. Firstly, applications deployed for extendeduse, i.e. ones that do not adopt an always on strategy (Section 2.7), rarely rely on justone of the strategies presented here. Secondly, almost all MAC protocols for rare eventsensing make use of component deactivation to the extend that they regularly turn offthe wireless transceiver.

The predominance of multi-strategy real-world deployments is reflected in the num-ber of citations of the surveyed papers. A the time of writing, the two most widely



cited articles1 included in this survey [Heinzelman et al. 2002; Ye et al. 2002] focus onsingle strategies: message suppression (Section 2.5) and component deactivation (Sec-tion 2.3) respectively. If the individual strategies are considered the building blocksof an efficient and successful WSN, which ones to use and how to combine them arefundamental considerations when developing real-world systems.

4. ASSOCIATED WORKIn situations where event occurrence cannot be detected at an individual node or by acollaborative evaluation by neighbouring nodes, sensed data can be transmitted to abase station for consolidated event determination. A simulated mine safety system [Liet al. 2008] takes this approach to detect gas leaks, water seepage and areas of highoxygen concentration that could provide refuge for mine workers in an emergency.Whilst the system as a whole detects rare events, the WSN nodes themselves performperiodic data capture only.

In addition to sensing geospatial phenomenon, malicious disruption of the WSNitself can be considered a rare event. [da Silva et al. 2005] propose a decentralisedthree phase rules based algorithm for intrusion detection focusing on multiple attacksstrategies including message delay, alteration and repetition. [Sun et al. 2007] and [Ro-man et al. 2006] describe the application of intrusion detection mechanisms to WSNswhilst [Czarlinska et al. 2007] consider attacks on sensor-actuators in hostile environ-ment to be rare events. Similarly, fault detection and tolerance in WSNs deployed todetect critical rare events [Mahapatro and Khilar 2013; Ould-Ahmed-Vall et al. 2012;Jurdak et al. 2011] are of significant importance.

Beyond dedicated WSNs, rare geospatial events can be monitored by collaborativeuse of personal electronic devices. Further, some classes of rare events can be predictedby off-line analysis of data captured from multiple heterogeneous sensor networks notoriginally deployed in a detection role.

Community Sensing [Krause et al. 2008] is an area of research investigating the useof the built-in sensors and communication capabilities of smart phones to detect andmonitor rare events. The iShake App [Dashti et al. 2011] and others [Faulkner et al.2011] aim to assist in the detection of earthquakes and similar rare events with mobilesensing devices. Feeds from social networking sites have also been evaluated for theirreal-time “social sensing” potential [Sakaki et al. 2010].

5. CONCLUSION AND FUTURE WORKEnergy efficiency has long been the focus of WSN research [Akyildiz et al. 2002; Yicket al. 2008]. A comparatively recent study [Anastasi et al. 2009] specifically surveysenergy conservation techniques for WSNs. For a given sensing task, innovative tech-niques that demonstrably reduce energy consumption while maintaining an appro-priate level of sensing functionality and network connectivity will continue to be indemand.

As an increasing number of wireless devices are deployed under the IoT banner,distinctions between network types (sensor, actuator, ad-hoc, body-area) will becomeblurred and more about deployment scenarios than equipment and protocols. ShouldIoT drive a convergence of underlying technologies and the standardisation of low cost,energy efficient, programmable wireless devices, multi-purpose networks rather thanpoint solutions may become the norm. Such a convergence of device use and networktraffic could cause a significant degradation in rare event detection delay unless the

1Citations as of October 7, 2015 - Google Scholar: [Heinzelman et al. 2002] 8,864 [Ye et al. 2002] 5,544.Microsoft Academic Search: [Heinzelman et al. 2002] 1,991 [Ye et al. 2002] 2,222.



system as a whole honours QoS rules requiring critical rare event notification mes-sages to be transmitted with high priority.

Regardless of the multi/single purpose nature of future deployments, it appears rea-sonable to predict the rare event strategise identified in this survey will continue tobe used in appropriate circumstances, yet the relative occurrence of each may changeover time; some becoming more or less significant based on the rise or fall in ubiquity ofothers. Collaboration may be more frequently used if energy harvesting techniques im-prove to the point where significant overpopulation of always-on devices are deployedto provide a more nuanced picture of rare event occurrences. Conversely, componentde-activation may become less prevalent if rechargeable battery and charging technol-ogy improves significantly in capacity and efficiency; the ability to store more energy,more quickly potentially negating the need to frequently shut down energy hungrycomponents.

Energy harvesting promises to provide extended life to WSNs and has been exten-sively investigated for data capture applications [Seah et al. 2009], but comparativelylittle research exists on the successful application of this technology to rare eventWSNs. An area suitable for further research is the impact unpredictable energy har-vesting patterns have on existing energy conservation techniques used to prolong thelife of battery powered rare event WSNs. Some schemes will prove unworkable yetothers may lend themselves to modifications that allow energy harvesting to furtherextend the life of the WSN whilst ensuring high detection probability and maintaininglow detection delay.

Attention could also be given to the challenges of providing low-latency forwarding ofevent notification messages where nodes collaboratively duty cycle to conserve energy.Existing opportunistic forwarding algorithms [Fußler et al. 2003; Sanchez et al. 2007]may prove unsuitable in certain rare event sensing scenarios.

REFERENCESIan F Akyildiz, Weilian Su, Yogesh Sankarasubramaniam, and Erdal Cayirci. 2002. A survey on sensor

networks. IEEE Communications Magazine 40, 8 (2002), 102–114.Kh Mahmudul Alam, Joarder Kamruzzaman, Gour Karmakar, and Manzur Murshed. 2012. Priority Sen-

sitive Event Detection in Hybrid Wireless Sensor Networks. In Proceedings of the 21st InternationalConference on Computer Communications and Networks (ICCCN). Munchen, Germany, 1–7.

Ahmed Ammar and Daryl Reynolds. 2015. An adaptive scheduling scheme for cooperative energy harvestingnetworks. Journal of Communications and Networks 17, 3 (2015), 256–264.

Giuseppe Anastasi, Marco Conti, Mario Di Francesco, and Andrea Passarella. 2009. Energy conservation inwireless sensor networks: A survey. Ad Hoc Networks 7, 3 (2009), 537–568.

Anish Arora, Prabal Dutta, Sandip Bapat, Vinod Kulathumani, Hongwei Zhang, Vinayak Naik, Vineet Mit-tal, Hui Cao, Murat Demirbas, Mohamed Gouda, and others. 2004. A line in the sand: a wireless sensornetwork for target detection, classification, and tracking. Computer Networks 46, 5 (2004), 605–634.

Sitaram Asur and Srinivasan Parthasarathy. 2007. Correlation-based Feature Partitioning for Rare EventDetection in Wireless Sensor Networks. In Proceedings of the 1st International Workshop on KnowledgeDiscovery from Sensor Data (Sensor-KDD). San Jose, CA, USA.

Luigi Atzori, Antonio Iera, and Giacomo Morabito. 2010. The internet of things: A survey. Computer networks54, 15 (2010), 2787–2805.

Seema Bandyopadhyay and Edward J Coyle. 2003. An energy efficient hierarchical clustering algorithm forwireless sensor networks. In Proceedings of the 22nd Annual Joint Conference of the IEEE Computerand Communications Societies (INFOCOM). IEEE, San Francisco, CA, USA, 1713–1723.

S Borzsony, Donald Kossmann, and Konrad Stocker. 2001. The skyline operator. In Proceedings of the 17thInternational Conference on Data Engineering. IEEE, Heidelberg, Germany, 421–430.

Richard R Brooks, Parameswaran Ramanathan, and Akbar M Sayeed. 2003. Distributed target classificationand tracking in sensor networks. Proceedings of the IEEE 91, 8 (2003), 1163–1171.

James Bucklew. 2013. Introduction to rare event simulation. Springer Science & Business Media.



Qing Cao, Tarek Abdelzaher, Tian He, and John Stankovic. 2005a. Towards optimal sleep scheduling insensor networks for rare-event detection. In Proceedings of the 4th International Symposium on Infor-mation Processing in Sensor Networks (IPSN). Los Angeles, CA, USA, 8 pages.

Qing Cao, Ting Yan, John Stankovic, and Tarek Abdelzaher. 2005b. Analysis of target detection performancefor wireless sensor networks. In Distributed Computing in Sensor Systems. Springer, 276–292.

Long Cheng, Chengdong Wu, Yunzhou Zhang, Hao Wu, Mengxin Li, and Carsten Maple. 2012. A surveyof localization in wireless sensor network. International Journal of Distributed Sensor Networks 2012(2012).

M.Y. Cheng, Y.B. Chen, and Winston K.G. Seah. 2013. Event-Driven Energy-Harvesting Wireless SensorNetwork for Structural Health Monitoring. In Proceedings of the 38th Annual IEEE Conference on LocalComputer Networks (LCN). Sydney, Australia, 9 pages.

Seong-Pyo Cheon, Sungshin Kim, So-Young Lee, and Chong-Bum Lee. 2009. Bayesian networks based rareevent prediction with sensor data. Knowledge-Based Systems 22, 5 (2009), 336–343.

Delphine Christin, Parag S Mogre, and Matthias Hollick. 2010. Survey on wireless sensor network tech-nologies for industrial automation: The security and quality of service perspectives. Future Internet 2,2 (2010), 96–125.

Alexandra Czarlinska, William Luh, and Deepa Kundur. 2007. Attacks on sensing in hostile wireless sensor-actuator environments. In Proceedings of the Global Telecommunications Conference (Globecom). IEEE,Washington, DC, USA, 1001–1005.

Ana Paula R da Silva, Marcelo HT Martins, Bruno PS Rocha, Antonio AF Loureiro, Linnyer B Ruiz, andHao Chi Wong. 2005. Decentralized intrusion detection in wireless sensor networks. In Proceedings ofthe 1st ACM international workshop on Quality of service & security in wireless and mobile networks.ACM, Montreal, Quebec, Canada, 16–23.

Sankar Narayan Das and Sudip Misra. 2015. Event-driven probabilistic topology management in sparsewireless sensor network. Wireless Sensor Systems 5, 4 (2015), 210–217.

Shideh Dashti, Jack Reilly, Jonathan D Bray, Alex Bayen, Steven Glaser, Ervasti Mari, and ProfessorsJonathan D Bray. 2011. iShake: Using Personal Devices to Deliver Rapid Semi-Qualitative EarthquakeShaking Information. GeoEngineering Report, Depart. of Civil and Environ. Engineering, Univ. of Cali-fornia, Berkeley, Feb 28 (2011).

Prabal Dutta, Mike Grimmer, Anish Arora, Steven Bibyk, and David Culler. 2005. Design of a wirelesssensor network platform for detecting rare, random, and ephemeral events. In Proceedings of the 4thinternational symposium on Information processing in sensor networks. IEEE Press, Los Angeles, Cali-fornia, USA, 70.

Jeremy Eric Elson. 2003. Time synchronization in wireless sensor networks. Ph.D. Dissertation. Universityof California, Los Angeles.

Xenofon Fafoutis and Nicola Dragoni. 2011. ODMAC: an on-demand MAC protocol for energy harvesting-wireless sensor networks. In Proceedings of the 8th ACM Symposium on Performance evaluation ofwireless ad hoc, sensor, and ubiquitous networks. ACM, Miami Beach, FL, USA, 49–56.

GaoJun Fan and ShiYao Jin. 2010. Coverage problem in wireless sensor network: A survey. Journal ofnetworks 5, 9 (2010), 1033–1040.

Matthew Faulkner, Michael Olson, Rishi Chandy, Jonathan Krause, K Mani Chandy, and Andreas Krause.2011. The next big one: Detecting earthquakes and other rare events from community-based sensors. InProceedings of the 10th International Conference on Information Processing in Sensor Networks (IPSN).IEEE, Chicago, IL, USA, 13–24.

Jessica Feng and Miodrag Potkonjak. 2002. Power minimization by separation of control and data radios.In Proceedings of the IEEE CAS Workshop on Wireless Communication and Networking, Vol. 9. IEEE,Pasadena, CA, USA, 112–121.

Holger Fußler, Jorg Widmer, Michael Kasemann, Martin Mauve, and Hannes Hartenstein. 2003.Contention-based forwarding for mobile ad hoc networks. Ad Hoc Networks 1, 4 (2003), 351–369.

Erol Gelenbe and EC-H Ngai. 2008. Adaptive qos routing for significant events in wireless sensor networks.In Proceedings of 5th IEEE International Conference on Mobile Ad Hoc and Sensor Systems. IEEE,Atlanta, GA, USA, 410–415.

James M Gilbert and Farooq Balouchi. 2008. Comparison of energy harvesting systems for wireless sensornetworks. International Journal of Automation and Computing 5, 4 (2008), 334–347.

Paul Glasserman, Philip Heidelberger, Perwez Shahabuddin, and Tim Zajic. 1999. Multilevel splitting forestimating rare event probabilities. Operations Research 47, 4 (1999), 585–600.

Peng Guo, Tao Jiang, Qian Zhang, and Kui Zhang. 2012. Sleep scheduling for critical event monitoring inwireless sensor networks. IEEE Transactions on Parallel and Distributed Systems 23, 2 (2012), 345–352.



Jose A Gutierrez, Marco Naeve, Ed Callaway, Monique Bourgeois, Vinay Mitter, and Bob Heile. 2001. IEEE802.15. 4: a developing standard for low-power low-cost wireless personal area networks. IEEE Network15, 5 (2001), 12–19.

Chih-Chieh Han, Ram Kumar, Roy Shea, and Mani Srivastava. 2005. Sensor network software update man-agement: a survey. International Journal of Network Management 15, 4 (2005), 283–294.

David C. Harrison, Winston K.G. Seah, and Ramesh Rayudu. 2015. Coverage Preservation in Energy Har-vesting Wireless Sensor Networks for Rare Events. In Proceedings of the 40th Annual IEEE Conferenceon Local Computer Networks”. IEEE, Clearwater Beach, USA.

Tian He, Sudha Krishnamurthy, Liqian Luo, Ting Yan, Lin Gu, Radu Stoleru, Gang Zhou, Qing Cao, PascalVicaire, John A Stankovic, and others. 2006. VigilNet: An integrated sensor network system for energy-efficient surveillance. ACM Transactions on Sensor Networks (TOSN) 2, 1 (2006), 1–38.

Tian He, Sudha Krishnamurthy, John A Stankovic, Tarek Abdelzaher, Liqian Luo, Radu Stoleru, Ting Yan,Lin Gu, Jonathan Hui, and Bruce Krogh. 2004. Energy-efficient surveillance system using wirelesssensor networks. In Proceedings of the 2nd international conference on mobile systems, applications,and services. ACM, Boston, MA, USA, 270–283.

Wendi B Heinzelman, Anantha P Chandrakasan, and Hari Balakrishnan. 2002. An application-specific pro-tocol architecture for wireless microsensor networks. IEEE Transactions on Wireless Communications1, 4 (2002), 660–670.

Chih-Fan Hsin and Mingyan Liu. 2004. Network coverage using low duty-cycled sensors: random & coordi-nated sleep algorithms. In Proceedings of the 3rd International Symposium on Information Processingin Sensor Networks (IPSN). Berkeley, CA, USA, 433–442.

Chalermek Intanagonwiwat, Deborah Estrin, Ramesh Govindan, and John Heidemann. 2002. Impact ofnetwork density on data aggregation in wireless sensor networks. In Proceedings of 22nd InternationalConference on Distributed Computing Systems. IEEE, Vienna, Austria, 457–458.

Kyle Jamieson, Hari Balakrishnan, and YC Tay. 2006. Sift: A MAC protocol for event-driven wireless sensornetworks. In Wireless Sensor Networks. Springer, 260–275.

Xiaofan Jiang, Joseph Polastre, and David Culler. 2005. Perpetual environmentally powered sensor net-works. In Proceedings of Fourth International Symposium on Information Processing in Sensor Net-works. IEEE, Los Angeles, CA, USA, 463–468.

Raja Jurdak, X Rosalind Wang, Oliver Obst, and Philip Valencia. 2011. Wireless sensor network anomalies:Diagnosis and detection strategies. In Intelligence-Based Systems Engineering. Springer, 309–325.

Yimei Kang, Yang Han, and Jiang Hu. 2012. A node Scheduling based on partition for WSN. In Proceedingsof Wireless Telecommunications Symposium (WTS). IEEE, London, United Kingdom, 1–6.

Aman Kansal, Jason Hsu, Sadaf Zahedi, and Mani B Srivastava. 2007. Power management in energy har-vesting sensor networks. ACM Transactions on Embedded Computing Systems (TECS) 6, 4 (2007), 32pages.

Krasimira Kapitanova, Sang H Son, and Kyoung-Don Kang. 2012. Using fuzzy logic for robust event detec-tion in wireless sensor networks. Ad Hoc Networks 10, 4 (2012), 709–722.

Heikki Karvonen, Juha Petajajarvi, Jari Iinatti, Matti Hamalainen, and Carlos Pomalaza-Raez. 2014. Ageneric wake-up radio based MAC protocol for energy efficient short range communication. In Proceed-ings of 25th Annual International Sympsoium on Personal, Indoor, and Mobile Radio Communications(PIMRC). 5 pages.

Vivek Katiyar, Narottam Chand, and Surender Soni. 2011. Clustering algorithms for heterogeneous wirelesssensor network: A survey. International Journal of Advanced Networking and Applications 2, 4 (2011),745–754.

S Kavitha and S Lalitha. 2014. Sleep Scheduling for Critical Eevent Monitoring in Wireless Sensor Net-works. International Journal of Advanced Research in Computer and Communication Engineering 3, 1(2014), 4974–4978.

Abtin Keshavarzian, Huang Lee, and Lakshmi Venkatraman. 2006. Wakeup scheduling in wireless sensornetworks. In Proceedings of the 7th ACM international symposium on Mobile ad hoc networking andcomputing. ACM, Florence, Italy, 322–333.

Andreas Krause, Eric Horvitz, Aman Kansal, and Feng Zhao. 2008. Toward community sensing. In Proceed-ings of the 7th international conference on Information processing in sensor networks. IEEE ComputerSociety, St. Louis, Missouri, USA, 481–492.

Bhaskar Krishnamachari and Sitharama Iyengar. 2004. Distributed Bayesian algorithms for fault-tolerantevent region detection in wireless sensor networks. IEEE Transactions on Networking 53, 3 (2004),241–250.



L Krishnamachari, Deborah Estrin, and Stephen Wicker. 2002. The impact of data aggregation in wirelesssensor networks. In Proceedings of 22nd International Conference on Distributed Computing SystemsWorkshops. IEEE, Vienna, Austria, 575–578.

Sunil Kulkarni, Aravind Iyer, and Catherine Rosenberg. 2006. An address-light, integrated MAC and rout-ing protocol for wireless sensor networks. IEEE/ACM Transactions on Networking 14, 4 (2006), 793–806.

Santosh Kumar, Anish Arora, and Ten-Hwang Lai. 2005. On the lifetime analysis of always-on wirelesssensor network applications. In Proceedings of 2nd IEEE International Conference on Mobile Adhocand Sensor Systems. IEEE, Washington, DC, USA, 3 pages.

Santosh Kumar, Ten H Lai, and Jozsef Balogh. 2004. On k-coverage in a mostly sleeping sensor network.In Proceedings of the 10th annual international conference on Mobile computing and networking. ACM,Philadelphia, PA, USA, 144–158.

Philip Levis, Sam Madden, Joseph Polastre, Robert Szewczyk, Kamin Whitehouse, Alec Woo, David Gay, Ja-son Hill, Matt Welsh, Eric Brewer, and others. 2005. TinyOS: An operating system for sensor networks.In Ambient intelligence. Springer, 115–148.

Huijiang Li, Neeraj Jaggi, and Biplab Sikdar. 2012. An analytical approach towards cooperative relayscheduling under partial state information. In Processdings of the 31st International Conference onComputer Communications (INFOCOM). IEEE, 2666–2670.

Mo Li, Yunhao Liu, and Lei Chen. 2008. Nonthreshold-based event detection for 3D environment monitoringin sensor networks. IEEE Transactions on Knowledge and Data Engineering 20, 12 (2008), 1699–1711.

Shuoqi Li, Sang H Son, and John A Stankovic. 2003. Event detection services using data service middlewarein distributed sensor networks. In Information Processing in Sensor Networks. Springer, 502–517.

Qilian Liang and Lingming Wang. 2005. Event detection in wireless sensor networks using fuzzy logic sys-tem. In Proceedings of the 2005 IEEE International Conference on Computational Intelligence for Home-land Security and Personal Safety (CIHSPS). IEEE, Orlando, FL, USA, 52–55.

Chong Liu, Kui Wu, Yang Xiao, and Bo Sun. 2006. Random coverage with guaranteed connectivity: jointscheduling for wireless sensor networks. Transactions on Parallel and Distributed Systems 17, 6 (2006),562–575.

Kay Soon Low, Win Nu Nu Win, and Meng Joo Er. 2005. Wireless sensor networks for industrial environ-ments. In Proceedings of International Conference on Computational Intelligence for Modelling, Controland Automation, and International Conference on Intelligent Agents, Web Technologies and InternetCommerce, Vol. 2. IEEE, Vienna, Austria, 271–276.

Xuanwen Luo, Ming Dong, and Yinlun Huang. 2006. On distributed fault-tolerant detection in wirelesssensor networks. IEEE Trans. Comput. 55, 1 (2006), 58–70.

Arunanshu Mahapatro and Pabitra Mohan Khilar. 2013. Fault diagnosis in wireless sensor networks: Asurvey. IEEE Communications Surveys & Tutorials 15, 4 (2013), 2000–2026.

Fernando Martincic and Loren Schwiebert. 2006. Distributed event detection in sensor networks. In Proceed-ings of International Conference on Systems and Networks Communications (ICSNC). IEEE, Papeete,Tahiti, French Polynesia, 43.

David Martins and Herve Guyennet. 2010. Wireless sensor network attacks and security mechanisms: Ashort survey. In Proceedings of 13th International Conference on Network-Based Information Systems(NBiS). IEEE, 313–320.

Prodromos-Vasileios Mekikis, George Athanasiou, and Carlo Fischione. 2013. A Wireless Sensor NetworkTestbed for Event Detection in Smart Homes. In Proceedings of International Conference on DistributedComputing in Sensor Systems (DCOSS). IEEE, Cambridge, MA, UAS, 321–322.

Bratislav Milic. 2012. Optimal stopping strategies in collaborative event detection in wireless sensor net-works. In Proceedings of 4th International Congress on Ultra Modern Telecommunications and ControlSystems and Workshops (ICUMT). IEEE, St. Petersburg, Russia, 842–849.

Sudip Misra, Satyadeep Mishra, and Manas Khatua. 2015. Social sensing-based duty cycle management formonitoring rare events in wireless sensor networks. IET Wireless Sensor Systems 5, 2 (2015), 68–75.

Prabhudutta Mohanty, Sangram Panigrahi, Nityananda Sarma, and Siddhartha Sankar Satapathy. 2010.Security Issues In Wireless Sensor Network Data Gathering Protocols: A Survey. Journal of Theoretical& Applied Information Technology 13 (2010).

Luca Mottola and Gian Pietro Picco. 2011. Programming wireless sensor networks: Fundamental conceptsand state of the art. ACM Computing Surveys (CSUR) 43, 3 (2011), 19.

Aziz Nasridinov, Sun-Young Ihm, Young-Sik Jeong, and Young-Ho Park. 2014. Event Detection in WirelessSensor Networks: Survey and Challenges. In Mobile, Ubiquitous, and Intelligent Computing. Springer,585–590.



Jonathan P Olds and Winston KG Seah. 2012. Design of an active radio frequency powered multi-hop wire-less sensor network. In Proceedings of 7th IEEE Conference on Industrial Electronics and Applications(ICIEA). IEEE, Singapore, 1721–1726.

ElMoustapha Ould-Ahmed-Vall, Bonnie Heck Ferri, and George F Riley. 2012. Distributed fault-tolerancefor event detection using heterogeneous wireless sensor networks. IEEE Transactions on Mobile Com-puting 11, 12 (2012), 1994–2007.

Nikolaos Pantazis, Stefanos A Nikolidakis, and Dimitrios D Vergados. 2013. Energy-efficient routing pro-tocols in wireless sensor networks: A survey. Communications Surveys & Tutorials, IEEE 15, 2 (2013),551–591.

Chulsung Park and Pai H Chou. 2006. Ambimax: Autonomous energy harvesting platform for multi-supplywireless sensor nodes. In Proceedings of 3rd Annual IEEE Communications Society on Sensor and AdHoc Communications and Networks, Vol. 1. IEEE, Reston, VA, USA, 168–177.

Gyuhae Park, Tajana Rosing, Michael D Todd, Charles R Farrar, and William Hodgkiss. 2008. Energyharvesting for structural health monitoring sensor networks. Journal of Infrastructure Systems 14, 1(2008), 64–79.

Joseph Polastre, Jason Hill, and David Culler. 2004. Versatile low power media access for wireless sensornetworks. In Proceedings of the 2nd international conference on embedded networked sensor systems.ACM, Baltimore, MD, USA, 95–107.

Kresimir Pripuzic, Hrvoje Belani, and Marin Vukovic. 2008. Early forest fire detection with sensor net-works: sliding window skylines approach. In Knowledge-Based Intelligent Information and EngineeringSystems. Springer, 725–732.

Venkatesh Rajendran, Katia Obraczka, and Jose Joaquin Garcia-Luna-Aceves. 2006. Energy-efficient,collision-free medium access control for wireless sensor networks. Wireless Networks 12, 1 (2006), 63–78.

Sidra Rashid, Usman Akram, Saad Qaisar, Shoab Ahmed Khan, and Emad Felemban. 2014. Wireless SensorNetwork for Distributed Event Detection Based on Machine Learning. In Proceedings of 2014 IEEEInternational Conference on Internet of Things (iThings 2014), Green Computing and Communications(GreenCom 2014), and Cyber-Physical-Social Computing (CPSCom 2014). IEEE, 540–545.

Joel JPC Rodrigues and Paulo ACS Neves. 2010. A survey on IP-Based wireless sensor network solutions.International Journal of Communication Systems 23, 8 (2010), 963–981.

Rodrigo Roman, Jianying Zhou, and Javier Lopez. 2006. Applying intrusion detection systems to wire-less sensor networks. In Proceedings of IEEE Consumer Communications and Networking Conference(CCNC). Las Vegas, Nevada, USA, 640–644.

Gerardo Rubino and Bruno Tuffin. 2009. Rare event simulation using Monte Carlo methods. John Wiley &Sons.

Takeshi Sakaki, Makoto Okazaki, and Yutaka Matsuo. 2010. Earthquake shakes Twitter users: real-timeevent detection by social sensors. In Proceedings of the 19th international conference on world wide web.ACM, Raleigh, North Carolina, USA, 851–860.

Juan Sanchez, Rafael Marin-Perez, Pedro M Ruiz, and others. 2007. BOSS: Beacon-less on demand strategyfor geographic routing inwireless sensor networks. In Proceedings of IEEE International Conference onMobile Adhoc and Sensor Systems. IEEE, Pisa, Italy, 10 pages.

Curt Schurgers and Mani B Srivastava. 2001. Energy efficient routing in wireless sensor networks. In Pro-ceedings of Military Communications Conference. Communications for Network-Centric Operations: Cre-ating the Information Force, Vol. 1. IEEE, Vienna, Virginia, USA, 357–361.

Winston KG Seah, Zhi Ang Eu, and Hwee-Pink Tan. 2009. Wireless sensor networks powered by ambientenergy harvesting (WSN-HEAP) - survey and challenges. In Proceedings of the 1st International Confer-ence on Wireless Communication, Vehicular Technology, Information Theory and Aerospace & ElectronicSystems Technology. IEEE, Aalborg, Denmark, 5 pages.

Jaydip Sen. 2010. A survey on wireless sensor network security. arXiv preprint arXiv:1011.1529 (2010).Megha Sharma and Ashutosh Kumar Singh. 2014. Analysis of Type-2 Fuzzy Systems for WSN: A Survey.

In Intelligent Computing, Networking, and Informatics. Springer, 1115–1123.Anmol Sheth, Kalyan Tejaswi, Prakshep Mehta, Chandresh Parekh, Rajul Bansal, S Merchant, T Singh,

Uday B Desai, Chandramohan A Thekkath, and K Toyama. 2005. Senslide: a sensor network basedlandslide prediction system. In Proceedings of the 3rd international conference on embedded networkedsensor systems. ACM, San Diego, CA, USA, 280–281.

Sujesha Sudevalayam and Purushottam Kulkarni. 2011. Energy harvesting sensor nodes: Survey and im-plications. IEEE Communications Surveys & Tutorials 13, 3 (2011), 443–461.

Bo Sun, Lawrence Osborne, Yang Xiao, and Sghaier Guizani. 2007. Intrusion detection techniques in mobilead hoc and wireless sensor networks. IEEE Wireless Communications 14, 5 (2007), 56–63.



Yanjun Sun, Shu Du, Omer Gurewitz, and David B Johnson. 2008a. DW-MAC: a low latency, energy efficientdemand-wakeup MAC protocol for wireless sensor networks. In Proceedings of the 9th ACM Interna-tional Symposium on Mobile Ad Hoc Networking and Computing. Hong Kong SAR, China, 53–62.

Yanjun Sun, Omer Gurewitz, and David B Johnson. 2008b. RI-MAC: a receiver-initiated asynchronous dutycycle MAC protocol for dynamic traffic loads in wireless sensor networks. In Proceedings of the 6th ACMConference on Embedded Networked Sensor Systems (SenSys). Raleigh, NC, USA, 1–14.

Harsh Sundani, Haoyue Li, Vijay Devabhaktuni, Mansoor Alam, and Prabir Bhattacharya. 2011. Wirelesssensor network simulators a survey and comparisons. International Journal of Computer Networks 2, 5(2011), 249–265.

QC Tang, QS He, MY Li, C Dong, DC Xu, and Xinxin Li. 2015. A self-powered wireless sensing node forevent-driven alerting based on a bi-stable vibration energy harvester. In Proceedings of 18th Interna-tional Conference on Solid-State Sensors, Actuators and Microsystems (TRANSDUCERS). IEEE, 1215–1218.

Kieu-Xuan Thuc and Koo Insoo. 2011. A collaborative event detection scheme using fuzzy logic in clusteredwireless sensor networks. AEU-International Journal of Electronics and Communications 65, 5 (2011),485–488.

Di Tian and Nicolas D Georganas. 2002. A coverage-preserving node scheduling scheme for large wirelesssensor networks. In Proceedings of the 1st ACM international workshop on wireless sensor networks andapplications. ACM, Atlanta, GA, USA, 32–41.

Sameer Tilak, Nael B Abu-Ghazaleh, and Wendi Heinzelman. 2002. A taxonomy of wireless micro-sensornetwork models. ACM SIGMOBILE Mobile Computing and Communications Review 6, 2 (2002), 28–36.

Daniel Tomicek, Yek-Hong Tham, Winston K.G. Seah, and Ramesh Rayudu. 2013. Vibration-powered Wire-less Sensor for Structural Monitoring during Earthquakes. In Proceedings of the 6th International Con-ference on Structural Health Monitoring of Intelligent Infrastructure. SHMII, Hong Kong, China, 10pages.

Tijs Van Dam and Koen Langendoen. 2003. An adaptive energy-efficient MAC protocol for wireless sensornetworks. In Proceedings of the 1st international conference on embedded networked sensor systems.ACM, Los Angeles, CA, USA, 171–180.

Federico Viani, Marco Salucci, Paolo Rocca, Giacomo Oliveri, and Andrea Massa. 2012. A multi-sensor wsnbackbone for museum monitoring and surveillance. In Proceedings of the 6th European ConferenceonAntennas and Propagation (EUCAP). IEEE, Prague, Czech Republic, 51–52.

Christopher M Vigorito, Deepak Ganesan, and Andrew G Barto. 2007. Adaptive control of duty cyclingin energy-harvesting wireless sensor networks. In Proceedings of 4th Annual IEEE CommunicationsSociety Conference on Sensor, Mesh and Ad Hoc Communications and Networks. IEEE, San Diego, Cal-ifornia, USA, 21–30.

Mehmet C Vuran and Ian F Akyildiz. 2006. Spatial correlation-based collaborative medium access controlin wireless sensor networks. IEEE/ACM Transactions onNetworking 14, 2 (2006), 316–329.

John Paul Walters, Zhengqiang Liang, Weisong Shi, and Vipin Chaudhary. 2007. Wireless sensor networksecurity: A survey. Security in distributed, grid, mobile, and pervasive computing 1 (2007), 367.

Geoffrey Werner-Allen, Konrad Lorincz, Mario Ruiz, Omar Marcillo, Jeff Johnson, Jonathan Lees, and MattWelsh. 2006. Deploying a wireless sensor network on an active volcano. Internet Computing, IEEE 10,2 (2006), 18–25.

Georg Wittenburg, Norman Dziengel, Stephan Adler, Zakaria Kasmi, Marco Ziegert, and Jochen Schiller.2012. Cooperative event detection in wireless sensor networks. Communications Magazine, IEEE 50,12 (2012), 124–131.

Georg Wittenburg, Norman Dziengel, Christian Wartenburger, and Jochen Schiller. 2010. A system for dis-tributed event detection in wireless sensor networks. In Proceedings of the 9th ACM/IEEE InternationalConference on Information Processing in Sensor Networks. ACM, Stockholm, Sweden, 94–104.

Lijie Xu, Jiannong Cao, Xuefeng Liu, Haipeng Dai, and Guihai Chen. 2012. EODS: An Energy-efficientOnline Decision Scheme in Delay-sensitive Sensor Networks for Rare-event Detection. In 18th Interna-tional Conference on Parallel and Distributed Systems (ICPADS). IEEE, 307–314.

Ning Xu, Sumit Rangwala, Krishna Kant Chintalapudi, Deepak Ganesan, Alan Broad, Ramesh Govindan,and Deborah Estrin. 2004. A wireless sensor network for structural monitoring. In Proceedings of the2nd international conference on Embedded networked sensor systems. ACM, Baltimore, MD, USA, 13–24.

Zhimin Yang, Kaijun Ren, and Chang Liu. 2013. Efficient data collection with spatial clustering in timeconstraint WSN applications. In Pervasive Computing and the Networked World. Springer, 728–742.

Yong Yao and Johannes Gehrke. 2003. Query Processing in Sensor Networks.. In Proceedings of the FirstBiennial Conference on Innovative Data Systems Research. Asilomar, CA, USA, 233–244.



Wei Ye, John Heidemann, and Deborah Estrin. 2002. An energy-efficient MAC protocol for wireless sensornetworks. In Proceedings of the 21st Annual Joint Conference of the IEEE Computer and Communica-tions Societies, Vol. 3. IEEE, New York, NY, USA, 1567–1576.

Wei Ye, John Heidemann, and Deborah Estrin. 2004. Medium access control with coordinated adaptivesleeping for wireless sensor networks. IEEE/ACM Transactions on Networking 12, 3 (2004), 493–506.

Jennifer Yick, Biswanath Mukherjee, and Dipak Ghosal. 2008. Wireless sensor network survey. Computernetworks 52, 12 (2008), 2292–2330.

Hongseok Yoo, Moonjoo Shim, and Dongkyun Kim. 2012. Dynamic duty-cycle scheduling schemes for energy-harvesting wireless sensor networks. IEE Communications Letters 16, 2 (2012), 202–204.

Mohamed-Haykel Zayani, Nadjib Aitsaadi, and Paul Muhlethaler. 2014. A new opportunistic routing schemein low duty-cycle WSNs for monitoring infrequent events. In Wireless Days (WD), 2014 IFIP. IEEE, 1–4.

Hongwei Zhang, Anish Arora, Young-ri Choi, and Mohamed G Gouda. 2007. Reliable bursty convergecast inwireless sensor networks. Computer Communications 30, 13 (2007), 2560–2576.

Junqi Zhang and Vijay Varadharajan. 2010. Wireless sensor network key management survey and taxon-omy. Journal of Network and Computer Applications 33, 2 (2010), 63–75.

Gang Zhou, Tian He, John Stankovic, Tarek Abdelzaher, and others. 2005. RID: radio interference detectionin wireless sensor networks. In Proceedings of 24th Annual Joint Conference of the IEEE Computer andCommunications Societies, Vol. 2. IEEE, 891–901.

Yanmin Zhu, Yunhuai Liu, and Lionel M Ni. 2012. Optimizing event detection in low duty-cycled sensornetworks. Wireless Networks 18, 3 (2012), 241–255.

Received June 2014; revised October 2015; accepted xxxx 201x


12 rare event detection and propagation in wireless sensor ......rare event detection and...

Documents